In the past, most of the opinion systems were either survey-based, in which the questionnaire decides what opinions are expressed, or if user-initiated, the user expresses his/her opinion in text and almost free format which mostly can lead to many opinions being expressed in multiple sentences, or paragraphs, making them hard to search, specifics mixed hidden in broad terms, difficult to analyze or compare, etc. In cases where a quantized method was used (e.g., numbered rating), the subject was pre-selected and then provided to the user, hence limiting the user's ability to very specifically select a subject.